University of Washington

Story Details

The Department of Mathematics at the University of Washington is one of the leading mathematics research departments in the United States. Several years ago, William Stein, associate professor of mathematics in the department, collaborated with more than 150 mathematicians from across the world to build an open-source computational software program called Sage. Sage can be used to solve problems involving a wide range of mathematics, including algebra, calculus, number theory, cryptography, numerical computation, group theory, combinatorics, graph theory, and more. Sage was developed to be a viable free open source alternative to software programs such as Magma, Maple, Mathematica, and Matlab.

Recently, Stein sought to expand the computational capabilities of Sage but felt limited by the hardware the department was using to run the software. “We were using a custom-built high-end server built by a local company, which was not nearly as powerful as what we needed,” says Stein. The department also wanted to increase the overall performance of Sage.

Additionally, the department wanted to make the automated testing of the platform faster and easier. Stability and rack space were major concerns, too. “We had some uptime issues with our hardware,” confirms Stein. “Also, in terms of space, we share server space with the entire mathematics department, and we were pushing the limits of how much space we were using for our servers.”

In November 2008, the department began to explore new server solutions to support Sage. After evaluating technology from Dell, IBM, and a local company, the department decided on a new server solution from Sun Microsystems. “The deal that Sun was able to offer us was much better than competitors' offers, and the technology itself seemed to fit our needs better,” says Stein. The solution includes four Sun Fire X4450 servers, powered by 2.6 Ghz Six-Core Intel Xeon processors, and one Sun Fire X4540 server that runs the Open Solaris Operating System. The X4540, which contains 24TB of disk space, is currently being used as a storage device for Sage. It provides the department with additional storage capacity in a compact device.

“The X4450 has multicore processing architecture and 128 gigabytes of memory, which were both really appealing to us,” Stein states. “Sun definitely put a lot of work into the engineering on this product, and it shows.” During the solution’s implementation, the mathematics department received SunSpectrum Silver support in addition to support from the Sun market development organization. “We have made use of this support several times and have been very satisfied with the results,” says Stein.

Since deploying the Sun solution in December 2008, the department has gained a huge increase in computing power. “Our previous infrastructure was fine, but the entire thing was only about one-third the capacity of a single Sun X4450 server,” remarks Stein. “We have much more power now than we did before, and Sage now runs computations much faster as well. The whole way I do mathematical calculations using Sage is so much different now that I have this hardware. I regularly set up computations that use the whole power of the machine at once, using parallel computing techniques, and I wouldn’t have been able to do that before.”

The Sun solution also speeds the automated testing of Sage. “Having a machine with so many cores makes it easy to do testing,” says Stein. “With a single-core machine, it would take me at least one hour to run the standard Sage test suite. With the Sun X4450, it only takes three minutes, because it runs all the tests in parallel.” System stability has also improved dramatically. “The uptime for this hardware is much better. We have had no mysterious system crashes in the six months since deploying,” says Stein.

Because the Sun X4450 is a space- and energy-efficient 2U size, the department can more easily address its server space limitations. “These servers are only half the height of the competition’s servers, so they only take up half the space in our server room,” remarks Stein. “That was really one of the critical reasons we chose the X4450.”